The data used in the simulation of land cover for the Jing-Jin-Ji region include climate data, topographic data, socioeconomic data, and the boundary data of a nature reserve. Climate data include observation data and scenario data. Observation climate data from 1961 to 2020 were collected from 176 meteorological stations located in the Jing-Jin-Ji region and its surrounding areas (Zhao et al., 2019). The climate scenario data included the scenarios of CMIP6 SSP1-2.6, SSP2-4.5, and SSP5-8.5 from 2020 to 2100, downloaded from the website: World Climate Research Programme(https://esgf-node.llnl.gov/projects/cmip6/). The SSP1- 2.6 represents a scenario under the influence of low mitigation pressure and low radiative forcing, and is also called a sustainable development scenario. The SSP2-4.5 represents a scenario of moderate radiative forcing, and maintains the current socioeconomic, scientific, and technological development trends. The SSP5-8.5 represents a scenario of high radiative forcing, of a high-speed development path dominated by fossil fuels (Zhang et al., 2019; Su et al., 2021).

The current land cover data for the Jing-Jin-Ji region in 2010 and 2020 were collected from the website: Resource and Environment Science and Data Center(https://www.resdc.n/) and classified into 11 types during the operation of the model. They were evergreen coniferous forest, deciduous coniferous forest, deciduous broad-leaved forest, mixed forest, scrubs, grassland, wetlands, cultivated land, built-up land, bare or sparse vegetation, and water bodies on the basis of distribution characteristics of land cover in Jing-Jin-Ji region and the classification principle of IGBP in land cover (Liu et al., 2009; Zhang et al., 2020). DEM data were collected from the SRTM data released by NASA(http://srtm.csi.cgiar.org/) with a resolution of 1 km × 1 km, and interpolated to 1 km × 1 km resolution by running the spatial interpolation method in ArcGIS. Boundary data from nature reserves were collected from the Ministry of Environmental Protection, Forestry Administration, National Bureau of Surveying and Mapping, and National Basic Geographic Information Center. High-resolution, remote-sensing-image data, vectorized by ArcGIS, were used (Yue et al., 2006; Fan et al., 2019).

The spatial-data accuracy of the three climate parameters, mean annual biotemperature (MAB), total annual precipitation (TAP), and potential evapotranspiration ratio (PER), directly influenced the reliability of the simulation results concerning land-cover scenarios (Fan et al., 2020). In order to ensure the simulation accuracy of the climate parameters, the Inverse Distance Weighted (IDW), Kriging, Spline, and High-Accuracy Surface Modeling (HASM) methods (Yue, 2011; Yue et al., 2006; 2007a; 2016; 2020; Fan et al., 2019), were respectively adopted to interpolate the MAB and TAP by integrating the longitude, latitude, and elevation data for the Jing-Jin-Ji region from 1991 to 2020 (T0). The simulation errors in TAP using the HASM, IDW, Kriging, and spline models are 26.21 mm, 107.52 mm, 51.79 mm, and 87.26 mm, respectively. The simulation errors in MAB are, respectively, 0.27°C, 0.98°C, 0.61°C, and 0.83°C. The results show that the accuracy of the HASM method for MAB and TAP is higher than that for the other models. Therefore, the HASM method is selected to interpolate the MAB and TAP data with a 1 km × 1 km resolution under the three scenarios, SSP1-2.6, SSP2-4.5, and SSP5-8.5, for the Jing-Jin-Ji region in 2020, 2040, 2070, and 2100. The PER data under the three scenarios are respectively calculated based on the interpolated data of MAB and TAP for 2020, 2040, 2070, and 2100.

Using key climate parameters, such as the annual average biological temperature, annual precipitation, and PER, an improved model of a natural-vegetation-ecosystem Holdridge Life Zone (HLZ) (Holdridge, 1947; Yue et al., 2005; 2020; Fan et al., 2019) was used to obtain the future scenarios of the natural vegetation ecosystem for 2020, 2040, 2070, and 2100, with a 1 km × 1 km resolution for the three scenarios, SSP1-2.6, SSP2-4.5, and SSP5-8.5, in the Jing-Jin-Ji region, respectively. Based on the characteristics of spatial similarity and consistency between the natural-vegetation-ecosystem type and the land-cover type in the spatial distribution, this paper constructed a land-cover scenario surface-modeling method (Scenario of Surface Modeling of Land Cover, SSMLC) driven by natural and human elements. This method combines the status quo of land cover and the quantitative factors of the spatial distribution ratios of various types, and comprehensively considered human factors such as traffic density, population density, and per capita GDP, as well as the driving impact of coordinated development planning of the Jing-Jin-Ji region, basic farmland on land-cover change and nature reserves and other policies and measures. The method includes the following three key technical steps.

2.3.1 Calculation of the corresponding probability between the natural-vegetation- ecosystem type and the land cover type

First, the input parameter form of the HLZ model is corrected from having discrete points to continuous spatial grid units to simulate the spatial distribution of the natural-vegetation- ecosystem types in different periods (Yue et al., 2007b). The modified model can be characterized as:

where BT(x,y,t), LR(x,y,t), and LP(x,y,t) are the base-2 logarithms of the annual average biological temperature, total annual precipitation, and PER at site $(x,y)$ for period $t$; LBT_i0, LRT_i0, and LPT_i0 are the standard values of the annual average biological temperature, annual precipitation, and annual PER at the spatial location represented by the i-th natural-vegetation-ecosystem type in the HLZ-model-discrimination system. When ${{D}_{j}}(x,y,t)=$ $min\{{{D}_{i}}(x,y,t)\}$, the value at site $(x,y)$ in period t is assigned to the j-th type of the natural-vegetation-ecosystem type. According to the SSP1-2.6, SSP2-4.5, and SSP5- 8.5 scenarios (with a resolution of 1 km × 1 km), the annual average biological temperature, average annual precipitation, and potential evapotranspiration data for the four periods of 2020, 2040, 2070, and 2100, the revised HLZ model of the natural vegetation ecosystem was run to obtain the spatial distribution data of the future scenarios of the natural vegetation ecosystem in the four periods of the above three scenarios.

Using the ArcGIS-spatial-overlay-analysis method, the land-cover type of each grid unit and the natural-vegetation-ecosystem type of the corresponding period are spatially overlapped; the similarity and consistency of their spatial distribution patterns are compared and analyzed. On this basis, the spatial data of the natural-vegetation-ecosystem type during the simulation period is added, and the corresponding probability of the natural-vegetation- ecosystem type and land-cover type in the spatial distribution is calculated. The theoretical formula for the specific calculation can be represented as:

where $HLZP{{(x,y)}_{k,t}}$ and $HLZP{{(x,y)}_{k,t+1}}$ characterize the probability that the natural- vegetation-ecosystem type at site (x, y) in the periods t and t+1 corresponds to the k-th land-cover type in period t; and $HL\_P{{(x,y)}_{k,t+1}}$ characterizes the comprehensive corresponding probability between the natural-vegetation-ecosystem type at site (x, y) and the k-th land-cover type in the periods t and t+1.

2.3.2 Spatial analysis model of the maximum probability of the land-cover change

In order to overcome the uncertain impact of large numerical differences among the population data, GDP data, traffic data, land-cover-type-distribution-probability data in period t, and natural-vegetation-ecosystem-distribution-probability data in period t+1 on model operation, the above data are normalized with mean value of 0 and variance of 1 respectively. The specific-spatial-normalization-calculation formula can be expressed as follows:

Where $\overline{P I(x, y)_{k, t}}, \overline{E I(x, y)_{k, t}}, \overline{T I(x, y)_{k, t}}$, and $\overline{L P(x, y)_{k, t}}$ are the normalized population- spatial-distribution-density coefficient, the per-capita-GDP-spatial-distribution-density coefficient, the traffic-accessibility-spatial-distribution-density coefficient corresponding to the k-th land-cover type, and the percentage-coefficient-of-land-cover-type distribution at site (x, y) in period t. x $\overline{H L_{-} P(x, y)_{k, t+1}}$. is the normalized-distribution-probability coefficient of the natural-vegetation-ecosystem type corresponding to the k-th land-cover type at site (x, y) in period t. $PN{{(x,y)}_{k,t}}$, $GDP{{(x,y)}_{k,t}}$, $TD{{(x,y)}_{k,t}}$, and $LP{{(x,y)}_{k,t}}$characterize the population-density value, traffic-accessibility value, per-capita-GDP value, and probability value of the land-cover-type distribution corresponding to class k land-cover type at site (x, y) in period t. $HL\_P{{(x,y)}_{k,t+1}}$characterizes the distribution-probability value of the natural-vegetation-ecosystem type corresponding to the k-th land-cover type at site (x, y) in period t. $\overline{P{{N}_{k,t}}},\ \overline{GD{{P}_{k,t}}},\ \overline{T{{D}_{k,t}}},\ \text{and}\ \overline{L{{P}_{k,t}}}$, characterize the population density, per-capita GDP, traffic accessibility, and mean value of the distribution probability of land-cover type distributed at site (x, y) in period t. PN(x, y)_k,t characterizes the mean value of the natural vegetation ecosystem at site (x, y) in period t. σ characterizes the overall standard deviation of the parameter.

To examine the different effects of human activity and policies on land-cover change in built-up areas, suburban areas, and rural areas in Jing-Jin-Ji, a spatial-analysis model that quantitatively identifies the maximum probability of land-cover change on a grid unit is constructed. The model uses geographically weighted regression (GWR) based on the normalized-population-distribution-density coefficient (PI), per-capita-GDP-density coefficient (EI), traffic-accessibility coefficient (TI), land-cover-type-distribution-probability coefficient (LP), and natural-vegetation-ecosystem-type-distribution-probability coefficient (HL_P). The theoretical formulas can be expressed as follows:

where $LC\_P{{({{x}_{i}},{{y}_{j}})}_{k,\ t+1}}$. characterizes the probability of type k-th land cover at site (x_i, y_j) in the period t+1; $\overline{L C_{-} T(x_{i}, y_{i})_{t+1}}$ characterizes the most likely land-cover type at site (x_i, y_j) in the period t+1; k = 1, 2, 3,..., 12 characterizes the numerical code of 11 land-cover types; α, β, γ, δ, ε and ω characterize the weight coefficients of each factor in the land-cover type, which use the k-nearest neighbor kernel function to obtain α + β + γ + δ + ε + ω = 1. The meaning of the other parameters is the same as the above.

2.3.3 Limiting rules for the impact of policy measures on land-cover change

In the context of a coordinated development strategy for the Jing-Jin-Ji region, the development and utilization of land resources must adhere to the premise of ecological priority and the principle of regional integration and coordinated development. Therefore, taking into account the relevant regulations and policies of the National Natural Package Reserve (NR) within the region, the Key Ecological Function Area (ER), which returns farmland to forest and grassland, basic farmland protection (BC), and the coordination-development strategy in Jing-Jin-Ji, the following restrictions on land-cover change have been formulated: A land-cover type in the NR area is prohibited from being converted to cultivated land and built-up land; a land-cover type in the ER area is prohibited from being converted to built-up land; a land-cover type with a slope (SLOPE) greater than 25 degrees is prohibited from being converted to agricultural land, and land-cover types in the BC area are prohibited from being converted to any land-cover type, other than cultivated land. The above restrictions can be expressed by the following logical judgment formulas:

where $LC\_T{{(x,y\text{)}}_{t}}$ characterizes the land-cover type at site (x, y) in the period t; $SLOPE(x,y)$ characterizes the slope at site (x, y) in the period t; and $Lcrop$ and $Lbuilt$ characterize cultivated land and built-up land.

In the process of simulating the spatial distribution of land-cover types in the t+1 period, formula (8) is first used to solve the possibility probability of various land-cover types in all the grid units of the whole study area in the t+1 period. Formula (9) is then used to identify the land-cover types with the maximum possibility probability of grid units (x_i, y_j) in the t+1 period. Combined with the logical judgment formula (10) of the land-cover type restriction rule, the maximum possible value of land-cover type for each grid unit (x_i, y_j) in the t+1 period is assigned and solved. In the process of using the discrimination conditions of the limiting rules, if the maximum value of formula (9) does not meet any of the conditions of formula (10), the type with the second possibility probability value of land-cover type in formula (9) shall be used for assignment, and so on, until the land-cover type of each grid unit (x_i, y_j) in the t+1 period is assigned, Finally, the spatial simulation analysis of land cover in the t+1 period for the whole study area is realized. In addition, in order to reduce uncertainty in the simulation results (arising from an over-long time scale for t simulation) after obtaining the simulation results for the land-cover spatial distribution in the t+1 period, the simulated land-cover data in the t+1 period are used as the starting and verification data for the t+2 period simulation; the above steps are then repeated, and then realize the simulation of land cover spatial distribution in t+2 period. In the simulation, the above methods and steps are repeated until the land-cover change simulation in all periods is realized.

In order to ensure the reliability of the simulation results, the overall accuracy evaluation method and Kappa coefficient verification method are used to compare and analyze the model simulation data with the current data concerning land cover in Beijing, Tianjin, and Hebei for 2020. The calculation method of simulation accuracy verification is as follows:

where P_ATA is the overall accuracy of the model simulation; KI is the Kappa coefficient of the model simulation; N is the total number of grids in the Jing-Jin-Ji region; k is the quantity of land-cover types (M = 12); p_ii is the number of grids whose simulated value of the i-th land-cover type is the same as that of the current data; p_i+ is the number of grids of the i-th land-cover type in the current situation data; and p_+i is the number of grids of the i-th land-cover type in the simulation data.

Simulation accuracy verification shows that the overall accuracy of the model for the land cover simulation in Jing-Jin-Ji in 2020 is 93.52%, and that of the Kappa coefficient is 93.07%. In order to verify the accuracy of each land-cover type simulated by the model, the simulation accuracy of the current value and simulation value of each land-cover type are calculated, respectively, based on the land-cover simulation data and current data of Jing-Jin-Ji in 2020 (Table 1). In Table 1, except for the simulation accuracy of deciduous broad-leaved forest, which is 87.90%, the simulation accuracy of the other land-cover types is higher than 90%. The above validation analysis shows that the model can effectively simulate the land-cover types in the Jing-Jin-Ji region.

By analyzing the simulation results for the Jing-Jin-Ji region (Figure 2-4) based on the three scenarios of SSP1-2.6, SSP2-4.5, and SSP5-8.5, we found that land cover shows the following distribution pattern: Evergreen coniferous forest is mainly distributed in mountainous areas 800-1500 m in the east of the Taihang Mountains; deciduous coniferous forest is mainly distributed in the north of the Taihang, constituting low mountains and hills of the northeastern Haihe Plain and some areas of the Yanshan Mountains; deciduous broad-leaved forest is mainly distributed in the Yanshan and northern Taihang mountainous areas; mixed forest is mainly distributed in the transitional zone between the northern Taihang and Yanshan mountains; and scrubs are widely distributed, mainly in mountainous areas 500 m above sea level. The grassland spatial distribution is the widest in the Jing-Jin-Ji region, with relatively continuous and concentrated distribution areas, in the 300-600 m, low-mountain areas and piedmont in the west of Handan, Xingtai, Shijiazhuang, and Baoding; at 600-1800 m above sea level in the middle of Zhangjiakou, and in the 400-1000 m low mountains and hills in Chengde, Tangshan, and Qinhuangdao. Water bodies are mainly distributed in lakes, rivers, reservoirs, and depressions within the Luanhe and Haihe river basins, while wetlands are mainly distributed in humid and low-lying areas around water bodies. Bare or sparse vegetation is mainly distributed in areas with little rain, in northwest Hebei, the desert area in Huailai County, and the saline-alkali zone around the Bohai Sea. Cultivated land is mainly distributed in Haihe Plain, in Yanshan’s low-mountainous areas and piedmont, and on the plateau of northwest Zhangjiakou. Built-up land is mainly distributed in the inland-river- system and plain areas with developed transportation in Jing-Jin-Ji.

Under the three scenarios of SSP1-2.6, SSP2-4.5, and SSP5-8.5, with continuously increasing temperature, precipitation, and human activity, especially against the background of the coordinated and integrated development of Jing-Jin-Ji, during 2020-2100, the simulated land-cover changes for the four periods of 2020, 2040, 2070, and 2100 in Jing-Jin-Ji show the following trends: Cultivated land shows a decreasing trend, and is mainly converted to built-up land and deciduous broad-leaved forest. Built-up land generally increases, especially in the vicinity of the Jing-Jin-Ji-land-transportation network, in the suburbs of existing urban built-up land, and in the coastal zone around the Bohai Sea and other regions. The land-cover types of evergreen coniferous forest, deciduous broad-leaved forest, deciduous coniferous forest, scrubs, and mixed forest generally increase. Driven by the policy of returning farmland to forest and grassland and the building of ecological civilization, some cultivated land with a slope over 25 degrees in the mountainous areas of the north of Jing-Jin-Ji is gradually converted to forest land. In addition, with increasing precipitation and implementation of desertification control measures in the north of Jing-Jin-Ji, the types of bare or sparse vegetation in the corresponding areas gradually shrink.

Statistical analysis of the simulation results for land-cover change in Jing-Jin-Ji under the SSP1-2.6 scenario (Figure 2 and Table 2) shows that the built-up land in Jing-Jin-Ji has the fastest growth rate in 2020-2100: Its area increases by 3011 km², with an average increase of 4.23% per decade. The area of evergreen coniferous forest, deciduous coniferous forest, deciduous broad-leaved forest, mixed forest, and scrub increases by 0.92%, 1.93%, 1.01%, 0.91% and 0.39%, respectively per decade. The area of grassland, wetland, cultivated land, water bodies, and bare or sparse vegetation shows a decreasing trend of 0.29%, 2.16%, 0.44%, 0.11%, and 1.55% respectively, per decade; the area of cultivated land decreases by 551 km² per decade from 2020 to 2100.

Statistical analysis of the simulation results for land-cover change in Jing-Jin-Ji under the SSP2-4.5 scenario (Figure 3 and Table 3) shows that the area of built-up land, evergreen coniferous forest, deciduous coniferous forest, deciduous broad-leaved forest, mixed forest, and scrubs in Jing-Jin-Ji continues to increase from 2020 to 2100, with an average increase of 4.56%, 1.02%, 2.08%, 1.52%, 2.96%, and 0.85% per decade. The area of grassland, wetlands, cultivated land, water bodies, bare or rare vegetation shows a continuous decreasing trend of 0.58%, 2.46%, 0.56%, 0.14%, and 1.87%, respectively, per decade. The area of cultivated land decreases by 639 km² per decade in the next 80 years, and the growth rate of built-up land is the fastest at 406 km² per decade.

Statistical analysis of the simulation results for land-cover change in the Jing-Jin-Ji region under the SSP5-8.5 scenario (Figure 4 and Table 4) shows that from 2020 to 2100, there will area of built-up land, evergreen coniferous forest, deciduous coniferous forest, deciduous broad-leaved forest, mixed forest, and scrubs continues to increase by 5.42%, 1.09%, 2.08%, 2.83%, 0.14%, and 1.65%, respectively, per decade. The area of grassland, wetlands, cultivated land, water bodies, bare or sparse vegetation continues to decrease by 0.60%, 3.29%, 0.88%, 0.54% and 3.46%, respectively, per decade. The area of cultivated land decreases the most: In the next 80 years, the area decreases by 1088 km² per decade, while the growth rate of built-up land is the fastest, with an average increase of 482 km² per decade.

The simulation results of land-cover change in Jing-Jin-Ji driven by SSP1-2.6, SSP2-4.5, and SSP5-8.5 (Figure 5) show that from 2020 to 2100, built-up land increases the fastest, with an average rate of 5.07% per decade. Wetlands decrease the fastest, with an average rate of 2.58% per decade. The intensity of land-cover change in the Jing-Jin-Ji region under the SSP5-8.5 scenario in 2020-2040, 2040-2070, and 2070-2100 is higher than that under the SSP1-2.6 and SSP2-4.5 scenarios. From 2020 to 2100, the land cover of the three scenarios will have the highest intensity of change in the period of 2020-2040 (average change per decade of 3.12%), and the lowest intensity of change in the period of 2070-2100 (average change per decade of 0.98%). During 2020-2040, the average dynamic change intensity of evergreen coniferous forest, deciduous coniferous forest, grassland, wetlands, cultivated land, built-up land, and water bodies per decade under the SSP1-2.6 scenario is 0.93%, 2.28%, 0.15%, 2.25%, 0.50%, 4.15% and 0.10% respectively, and the above change intensity is lower than that under the SSP2-4.5 and SSP5-8.5 scenarios. In all the land-cover types during 2040-2070, the dynamic change in degree of scrubs under SSP2-4.5 is the lowest (the average change is only 0.01% per decade), and the dynamic change degree of bare or sparse vegetation under SSP5-8.5 is the highest (the average change is 2.75% per decade). In all the land-cover types during 2070-2100, the dynamic change in degree of water bodies under the SSP1-2.6 scenario is the smallest (the average change in degree per decade is only 0.03%), while the dynamic change in degree of mixed forest under the wet SSP5-8.5 scenario is the highest (the average change in degree per decade is 2.32%). In short, driven by the coupling of natural climate change and human activity, the change intensity of land cover in Jing-Jin-Ji during 2020-2100 under the SSP5-8.5 scenario is higher than that under the SSP1-2.6 and SSP2-4.5 scenarios.

The land cover scenario surface-modeling method (SSMLC) (Fan et al., 2005; 2015; Yue et al., 2007b) aims to expand and modify the SMLC model driven by natural climate change, in view of the defects in the existing I-O, IMAGE, cloud, CA, SD, and flux models for land-cover simulation (Alcamo et al., 1994; Clarke et al., 1997; Verburg et al., 1999; 2002; Oshiaki IchinoSe, 2003; He et al., 2005; Fan et al., 2015; Liu et al., 2017). Then, a spatial-simulation method for medium- and large-scale land cover scenarios, which simultaneously takes into account the driving effects of natural climate and human activity on land-cover change, is developed (Holdridge, 1947; Yue et al., 2005). The main advantages of the SSMLC model include: It overcomes the defects of the I-O model. The I-O model focuses on the economic utility analysis of land-cover change and ignores the driving impact of climate change on land-cover change. It overcomes the limitations of the IMAGE model, which mainly focuses on the driving of agricultural ecological processes, ignores the driving effect of natural and human factors on land-cover change from the ecological pattern, and is poor at obtaining the ecological process parameters of a large area. It makes up for the model defects of the CLUE model in the simulation process.

The CLUE model must first constrain the number of future land-cover types under various scenarios in order to realize the maximum probability simulation of the spatial distribution of various land-cover types under various scenarios. It overcomes the shortcomings of the SD model. The SD-model-driving-cycle structure is too complex, especially on large regional scales, and it experiences difficulty in distinguishing among the different driving effects affecting elements. It effectively makes up for the defects of the FLUS model. The FLUS model is mostly used for down-scaling simulations of cultivated land and various types of urban land, where human activity has a strong interference effect on large regional scales. In short, the spatial results of the SSMLC model in the Jing-Jin-Ji region show that this method ably takes into account the driving effects of natural climate elements, human factors, and policy measures on land-cover change; and that this method successfully simulates future scenarios concerning land-cover change in the Jing-Jin-Ji region. In addition, in the spatial simulation of land-cover change for each period in the future under each scenario, the results of the previous period can be used as the basic data for simulation of the next period. This mitigates the uncertainty caused by cumulative error arising from simulations over a large time scale, and ensures the model can be used for the simulation of land-cover change over such timescales. However, because the current SSMLC model mainly simulates the spatial and temporal and changes in land-cover category driven by natural climate and human factors, the current simulation is robust regarding the internal structure of built-up land and cultivated land that has a strong interference with human activity. There are still some limitations in portraying. In future work, we will draw on the existing model to analyze the structural conversion mechanism of various built-up lands and cultivated lands, adding to and improving the model.

Based on the SSMLC model, the simulation results of future land-cover changes in the Jing- Jin-Ji region under the three scenarios of SSP1-2.6, SSP2-4.5, and SSP5-8.5 show that: From 2020 to 2040, the land-cover change in the Jing-Jin-Ji region is in a period of high change intensity, at 3.11% per decade. After entering 2040, the rate of change gradually slows down. The intensity of change per decade for the two periods of 2070 and 2100 is 1.36% and 0.98%, respectively. This simulation result confirms that in the “Jing-Jin-Ji Coordinated Development of Land Use Master Plan (2015-2020),” “Jing-Jin-Ji Urban Agglomeration Development Plan,” “Hebei Xiong’an New District Built-up Plan,” “Beijing Urban Planning 2016”—under the guidance of the development plans for the Jing-Jin-Ji region such as 2035, Tianjin Urban Planning (2015-2030) and Hebei Urban System Planning (2016-2030), the land-cover change in this area will be in a high-intensity change between 2020 and 2040, especially built-up land will reach an annual growth rate of 0.58%, while the area of cultivated land will decrease at an annual rate of 0.06%. At the same time, in the Jing-Jin-Ji region, driven by policies that consolidate the effects of poverty alleviation, the building-up of ecological protection forests, and the increase in average temperature and precipitation (Zhao et al., 2019), the mountainous forest area shows a rapidly increasing trend. In addition, due to the rapid progress in urbanization of the Jing-Jin-Ji region, the development of various nodes in the Jing-Jin-Ji metropolitan area, and the rapid rise in population (Fan, 2008; Fang, 2017; Chen et al., 2018; Ouyang and Zhu, 2020), eco-tourism has developed rapidly (Fang et al., 2017). This will cause a rapid increase in water demand, which will result in a reduced water supply in an area that already has a water shortage; there will be a concomitant reduction in the area of wetlands and water bodies.

Via a combination of different, shared socio-economic paths (SSPs) and typical concentration paths (RCPs) (Zhang et al., 2019; Su et al., 2021), the SSP1-2.6 scenario is a sustainable development scenario. Future spatiotemporal changes in land cover in the Jing-Jin-Ji region simulated under the SSP1-2.6 scenario are closer to being driven by the coordinated development of the Jing-Jin-Ji region and various land-resource-development policies and ecological built-up measures. The SSP2-4.5 scenario is a development model that continues the historical trend and the status quo. In the process of land-cover change, the importance of ecological protection and the sustainable use of resources is lower than that of the SSP1-2.6 scenario. Therefore, the intensity of future land-cover changes in this scenario will be higher. The SSP5-8.5 scenario, as a development path with rapid economic development at its core and one that underestimates ecological sustainability, will lead to a development trend of land-cover change in the Jing-Jin-Ji region that is higher than in the other two scenarios. In particular, built-up land has been increasing rapidly and the area of cultivated land has been greatly reduced.

As the strategic core area of national economic development and the main area of new urbanization (Fan, 2008; Fang, 2017; Chen et al., 2018; Ouyang and Zhu, 2020), Jing-Jin-Ji megacity agglomeration will not only occupy a large amount of cultivated land (Zhao, 2016), but will cause a corresponding urban-heat-island effect (Yue et al., 2019). Therefore, in the process of the coordinated development of Jing-Jin-Ji, it is necessary to comprehensively analyze the interaction and coupling effects of the Jing-Jin-Ji megacity cluster and the eco-environment (Fang et al., 2017; Ouyang and Zhu, 2020), and to seek the best sustainable development model. For example, with the rapid urbanization of Jing-Jin-Ji, the urbanization rate will reach 82% by 2050 (Gu et al., 2017), and the urban population will reach 118 million (Wang et al., 2016; Li and Kuang, 2019). This will increase the existing pressure on the use of water resources in the Jing-Jin-Ji region, and reduce and shrink the area of water bodies and wetlands. According to the simulation results, in the process of the coordinated development of Jing-Jin-Ji, it is necessary to strictly control the sustainable development and utilization of land resources, so as to achieve an optimal balance between rapid socioeconomic development and eco-environmental protection (Fang, 2017; Ouyang and Zhu, 2020). First, we must strictly control the occupation of cultivated land resources by built-up land, and curb the rapid decline in cultivated land during rapid urban development. Second, we must scientifically plan the sustainable use of ecological land in the Jing-Jin-Ji region, especially in the Jing-Jin-Ji urban agglomeration. When vigorously developing eco-tourism in the suburbs and mountainous areas of the Jing-Jin-Ji region, it is necessary to strengthen the protection of water resources and wetland resources and to establish prohibited development zones to curb the decline in water bodies and wetlands. In short, in the process of coordinated development in Beijing, Tianjin, and Hebei, these separate areas should be regarded as a whole for urban-development planning, industrial-structure adjustment and layout, and ecological-environment-protection planning. This would facilitate the scientific, rational development and utilization of land resources in the Beijing-Tianjin-Hebei region, and therefore reduce the unreasonable interference intensity arising from human factors.